No, i had no luck. But i’m glad i started this topics, as it gains attention from others - maybe nvidia would fix it in the end. This is really frustrating experience at the moment using nvidia on linux
1 Like
Yeah thanks for opening this, there is a similar thread on this forum. Can you confirm whether you’re alse getting:
nvidia-modeset: ERROR: GPU:0: Display engine push buffer channel allocation failed: 0x65 (Call timed out [NV_ERR_TIMEOUT])
when you resume from suspend?
at the moment i’m using prime-select nvidia only, so no, looks like i’m getting different kind of error
UPD. i read through your thread - yes, symptoms look quite similar. I can’t confirm if the error on my side is the same as yours - have no time for experimenting right now
Sadly though that Linux “desktop” users are lowest priority in means of support for Nvidia
My solution was slightly different but as I was having similar issues (Ubuntu 20.04 crashing on suspend) I thought I’d share the solution here in case it helps anyone else.
In my case I use CUDA quite a lot and I had installed the driver package from the NVIDIA CUDA repository
This package appears to not setup the systemd services correctly for suspend and resume. The solution was:
- comment out the nvidia developer repositories in
/etc/apt/sources.list
sudo apt purge nvidia-driver-*
sudo apt-add-repository ppa:graphics-drivers/ppa
sudo apt install nvidia-driver-<version> # currently 470
sudo shutdown -r now
I suspect there may be a more elegant solution using apt’s options to only update the drivers from ppa:graphics-drivers and not the nvidia developer repository.
In the mean time I’ll see if I can log a bug with nvidia.
1 Like
Should it be nvidia-driver-* instead?
1 Like
So, are you saying you’ve got hibernate working properly using drivers version 470 from ppa ?
So, are you saying you’ve got hibernate working properly using drivers version 470 from ppa ?
Hibernate doesn’t work as I am using secure boot and linux doesn’t support suspend to disk for systems running secure boot (see kernel lockdown and related features). This is nothing to do with Nvidia.
I have suspend to RAM working correctly on both desktop and laptop systems running 470.
I’m not using secure boot and i’ve got suspend to RAM working since 455 or before. I’ve started this thread about suspend to disk problem
May be useful here - solution with blank screen on resume after suspend:
I had similar problems and created the following guide. For me the steps in the guide solved all the problems related to hibernation and boot. How to install Nvidia 470.57.02 Drivers for Ubuntu 20.04 using a run file · GitHub
WIth PreserveVideoMemoryAllocations enabled, I get the error:
failed to quiesce async: error -5
Without it enabled, hibernate works less than half of the time. Suspend seems to be just as unreliable… I’m running driver version “470.57.02”, the next version “470.63.01” doesn’t have anything in the release notes regarding power management so I assume nothing is fixed there either.
It seems power management with the Nvidia drivers became less reliable for me around 46x.
Are there any plans for this to be fixed? For now I’m working around the issue by running in Intel only mode.
1 Like
I had a similiar issue withOpenSuSE Leap 15.3. (PC with two Nvidia GT 710). Unfortunately, patching nvidia-sleep.sh didn’t work for me.
However, disabling the nvidia suspend/resume services, i.e. going back to kernel callback, solved my problem (460 was fine anyway, 470 works now) …
Switched workstation to Windows, and servers to have Intel cards. Problem solved.